> It’s a bit complicated to set up […] it’s really fast

Sigh, just use Perl. Writing code with the general regex engine took me only one minute of effort, but it runs already nearly 500× faster than codeplea's optimised special purpose code.

Why 100000 loops and not 10 like in the original code? Otherwise Benchmark.pm will show "(warning: too few iterations for a reliable count)".

----

benchmark.php 100000 loops:

    Loaded 3000 keywords to search on a text of 19377 characters.

    Searching with aho corasick...
    time: 329.3522541523

----

benchmark.pl 100000 loops:

    Benchmark: timing 100000 iterations of regex...
    regex: 0.691561 wallclock secs ( 0.69 usr +  0.00 sys =  0.69 CPU) @ 144927.54/s (n=100000)

----

benchmark.pl (fill in the abbreviated ... parts from benchmark_setup.php):

    #!/usr/bin/env perl
    use Benchmark qw(timethese :hireswallclock);
    require Time::HiRes;
    my @needles = qw(
    abandonment abashed abashments abduction ...
    );
    my $haystack = 'unscathed grampus ...
    heroically';
    my $n = join '|', @needles;
    timethese 100000, {
        regex => sub {
            my @found;
            while ($haystack =~ /($n)/cg) {
                push @found, [$1, pos $haystack];
            }
            return @found;
        },
        index => sub {
            my @found;
            for (@needles) {
                my $pos = index $haystack, $_;
                push @found, [$_, $pos] if -1 < $pos;
            }
            return @found;
        }
    };